Last sunday, I read an article of J. Chem. Informatics.
The title was “Bioalerts: a python library for the derivation of structural alerts from bioactivity and toxicity data sets”.
URL is following
The author developed python library that analyse relation ship between structure and tox / bioactivity.
Sometime I use machine-learning or knowledge based classification methods to design molecule. ML is often tendency to be black box, and knowledge based methods are depend on data we have.
Bioalerts uses statistical method based on substructure based on morgan fp of training molecules.
And the library provides functionalities for the easy visualization of the derived structural alerts.
It useful for chemist, because the library provides substructure with probability. I think it’s easy to understand.
Fortunately, all source code and document can get from github. ;-)
I was interested the library because, the library use students’s t test or komogorov sirmov test depends of dataset.
It’s important for objective judgement.
BTW, If I use the library in my project, I need lots data points for training.
So, I think the library is suitable for general target i.e. ADMET or late stage of project.
Hmm… difficult problem for me.